Soundtrack: "For What It’s Worth" by Buffalo Springfield.
- Paranoia strikes deep
- Into your heart it will creep
- It starts when you’re always afraid…
My Bigger Deal colleague, Richard Whitehouse, his heart in his mouth, writes that the Sky is Falling and that online poker as we know it is about to end. Or at least there’s some danger of that. The source of his hand-wringing is the reports about a series of matches played by poker pro’s Phil Laak and Ali Eslami against "Polaris" – the most recent in a series of poker playing programs to come out of an artificial intelligence poker research group at the University of Alberta.
The bad news is that while Laak and Eslami won, it was only barely – what any poker-savvy observer would call a draw.
There were some interesting features to the matches (including playing a "duplicate" format to ensure that luck didn’t factor into the results), but that ground has been well-covered in the various reports linked here.
I’m fairly sure it won’t – read on.
For a bot to consistently win money in online poker it has to do two things:
- Beat the other players
- Avoid getting detected and busted by the online sites, such as PokerStars
Let’s consider problem #1 – beating the players. That’s easy; it’s happening already, and will only get worse.
I have a good friend (and former colleague at PokerStars), Terrence Chan, who is a fairly well-known pro poker player. Terrence came second to Hoyt Corkins in the $2500 buy-in No-Limit Hold’em event at the WSOP this year, but he makes his living (a good one, thank you) playing cash games, most of them on the Internet. In his extremely interesting blog, Terrence writes:
Of course, it’s totally obvious that one day limit hold’em will be solved by a computer. I have no real good guesses as to when that will happen, or when they will overtake the best humans as the best players in the world (we don’t know how close to optimal the best players in the world play).
Just so. It’s a question of when, not if. And already there are bots that can routinely beat most of the players in the game – Terrence is simply wondering when they’ll beat the best players. After all, folks, this game is not rocket science. Compared to chess, it’s tiddly winks. And please don’t start with me about the personal element, bluffing, yada yada. Computers know perfectly well how to bluff, when to bluff, etc. You just wire an RNG (random number generator) to decide when to bluff, based on game theoretically correct bluffing percentages. Child’s play – and there’s no pattern to it.
Whether a computer will ever be the best player in no-limit hold’em is perhaps a slightly more complex issue, because the science/art equation tips a bit more toward art (i.e. playing your opponent versus playing the cards). But some bots already do a pretty fair job of beating big-bet poker. Furthermore, I’m sure that even world class players have betting patterns that they might not even recognize, but a good artificial intelligence algorithm would.
In short, some day in the future, a computer program will certainly be the best limit hold’em player in the world. If I had to bet, I’d wager that one day a computer program will be the best no-limit hold’em player in the world.
But as I pointed out above, that only gets the bot-masters halfway there. They also have to get past the online sites.
Now, let me pose a rhetorical question: Do you suppose that the poker sites will allow bots to destroy the industry that they’ve worked so hard to create, and that has brought them billions (with a ‘b’) of dollars?
Yeah, that’s kinda what I thought, too.
As many of you are aware, up until April 1st of this year, I was the poker room manager at PokerStars.com. I saw the first bots appear at the edge of the forest, make their way across the field, and storm the castle. I watched the battle rage between the bots and the security guys on our side.
A few bots have made some money, but mostly they’ve gotten caught before they could do any serious damage. One of the reasons is that to really leverage a bot, you need to get a bunch of copies of that bot running. When you get lots of them running, patterns emerge. All kinds of patterns. Some of that has been discussed in other forums, such as 2+2, but I’m not going to elaborate here.
In general, the bot detection technology has improved dramatically over the past couple of years, and it’s getting better daily. It’s in the poker sites’ interest to be as bot-free as possible, and they’re working hard at it. Of course, the bot-writers will improve their technology too, but right now, I think the online sites are winning the war, and in fact are increasing the gap.
Also, running a bot causes you to risk a lot of money. For instance, if you have a bot that’s beating $30-60 limit hold’em, it will need a bankroll of $15,000-$20,000 to be safe (remember, it’s a bot, not a magician). If the site on which you’re running this bot catches you, you’re out $18,000; that’s a big hit on your bankroll. And if you’re running 3-4 bots and they catch all of them, then you’re out $50K or whatever. And you can’t win that $50K back because you’re persona non grata at the poker site.
Now, will online poker ever be completely bot-free? Probably not. Particularly if the bot is used as an "assistant" to provide statistics, suggest plays, and/or auto-fold bad hands, they will be difficult (I didn’t say impossible) to detect. But that’s a very different scenario than the one Richard Whitehouse fears: an army of cyborgs gobbling up the unwitting carbon-based online players.
So, that’s the situation: the press will serve up stories about machines beating people at poker – it makes good copy, feeds our fears of being replaced by robots, and gives them yet another excuse to show us a picture of Jennifer Tilly. But as far as I can see into the future, in the real day-to-day world of online poker, a good player who sticks to his or her A-game will take home the money.
